ggplot2: Prevent geom_bar from being alphabetical, and y scale not showing breaks

Question

I have two issues with my ggplot.

Data:

> dput(cts2)
structure(list(country_name = c("United States", "Canada", "India", 
"Bots", "Estonia", "Mexico", "Portugal", "Finland", "United Kingdom", 
"New Zealand", "Australia", "Russia", "Denmark", "Sweden", "Poland", 
"Ireland", "Ghana", "Netherlands", "Chile", "Other"), freq = c(716288L, 
77290L, 14925L, 12393L, 9526L, 9307L, 9003L, 6733L, 6174L, 3860L, 
3706L, 3553L, 3444L, 2371L, 1768L, 1532L, 523L, 286L, 234L, 909L
)), row.names = c(54L, 9L, 24L, 55L, 14L, 32L, 39L, 15L, 53L, 
34L, 3L, 44L, 13L, 49L, 38L, 25L, 19L, 33L, 10L, 27L), class = "data.frame")

As you can see, the data is already in the order that I want it. Highest to Lowest, with the exception of "Other" which should be at the end. Now when I try a simple ggplot with geom_bar:

ggplot(cts2, aes(x=factor(country_name), y=freq)) +
  geom_bar(stat="identity")+ 
  scale_y_continuous(trans='sqrt', labels = scales::comma, breaks = c(400, 10000, 100000, 700000))+
  theme(axis.text.x = element_text(angle = 90, hjust = 1))

The result has two major issues.

The bars are in alphabetical order. When looking at other questions about this issue, the answers were always 'use stat="identity"' or 'use a factor'. Well, as you can see, I have both of those with no effect.
The Y scale does not show the level of 400. 400 itself is arbitrary, as I want to add a lot breaks. The final barchart will be a large image, so there is space for more detail.

In regards to the Y scale, even with the trans='sqrt', The United States data point really makes it difficult to see the differences between the rest of the data items. Is there a better way to visualize this? Log2 and Log10 both looked even worse.

score 1 · Accepted Answer · edited Dec 06 '18 at 19:14

1

You need to set the factor levels. My guess is that setting a 400 label doesn't leave enough room to print the label without going negative. Negative numbers can't exist under a sqrt transform.

ggplot(data = cts2, aes(x=factor(country_name, levels = c(unique(cts2$country_name))), y=freq)) +
  geom_bar(stat="identity")+ 
  scale_y_continuous(trans='sqrt', labels = scales::comma, breaks = c(100, 10000, 100000, 700000))+
  theme(axis.text.x = element_text(angle = 90, hjust = 1))

See: Using ggplot2, can I insert a break in the axis? for some ideas on how to deal with such a large gap. In particular, gap.barplot from the library plotrix will allow you to place a gap in your bar chart.

Another approach suggested there is a simple table. You can use grid.table in the library gridExtra to make a nice looking table from the dataframe.

edited Dec 06 '18 at 19:14

Jared C

362
7
19

answered Dec 06 '18 at 18:13

jsta

3,216
25
35

Great, this solves issue #1. And #2? The Y scale is a big problem. – Jared C Dec 06 '18 at 18:15
I am not looking for a way to take out the white space on the Y axis. I just want to figure out why y axis labels aren't all showing up (for example the 400) and how to solve that. – Jared C Dec 06 '18 at 18:17
@JaredC You don't like `scale_y_log10()`? (Without `trans = 'sqrt'`). – Rui Barradas Dec 06 '18 at 18:55
`scale_y_log10()` looks much worse with this data. – Jared C Dec 06 '18 at 18:58

ggplot2: Prevent geom_bar from being alphabetical, and y scale not showing breaks

1 Answers1